5  Designing the Course Feeder Workbook

This chapter defines the standard feeder Excel workbook used to generate a complete Quarto-based course website. The workbook separates course content, delivery logistics, and grading structure into three clearly defined sheets.

This separation is intentional:

5.1 Workbook Overview

The feeder workbook must contain three sheets:

Sheet Name Purpose
course_details What is taught (modules, lectures, topics)
sections How and when it is taught
grading How students are evaluated

Only the first sheet is typically AI-generated. The other two are usually human-authored (or lightly assisted).

5.2 Sheet 1 — course_details (Core Content Sheet)

5.2.1 Purpose

This sheet defines the academic structure of the course:

  • Modules
  • Lectures
  • Titles
  • Descriptions
  • Delivery modality (online / on-campus)

This is the only sheet required to generate:

  • Lecture Notes (LN)
  • Presentations (P)
  • Labs
  • Highlights
  • Wrap-ups
  • Syllabus lecture tables

5.2.2 Required Columns

The course_details sheet must contain exactly the following columns:

Column Description
Modules Module identifier (M00–M07)
Lecture Lecture identifier (L01–L16)
Title Lecture title
Description One–two paragraph description
Online TRUE / FALSE
On-Campus TRUE / FALSE

At least one of Online or On-Campus must be TRUE.

5.2.3 Blank Template (Instructor-Facing)

This sheet is generated blank by default:

Modules Lecture Title Description Online On-Campus
M00 L01
M00 L02
M01 L03
… …

This ensures:

  • No accidental AI hallucinations
  • Full instructor control
  • Easy manual editing

5.2.4 Optional AI Population (OpenAI)

Once the structure exists, instructors may populate Title and Description using AI.

5.2.5 .env file

  • Create a .env file in the root directory of the project
  • Add the following line to the file:
OPENAI_API_KEY=your_api_key_here
  • Replace your_api_key_here with your actual OpenAI API key

OpenAI API Key
import openai
import pandas as pd

openai.api_key = os.getenv("OPENAI_API_KEY")

def generate_lecture(title_seed):
    prompt = f"""
    Generate a lecture title and a two-sentence description
    for a graduate business analytics course.

    Topic seed: {title_seed}
    """

    response = openai.ChatCompletion.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": prompt}],
        temperature=0.4
    )

    return response.choices[0].message.content

df = pd.read_excel("Course-Schedule.xlsx", sheet_name="course_details")
df.loc[0, "Title"] = "Introduction to Generative AI in Business"
df.loc[0, "Description"] = generate_lecture("Generative AI foundations")

df.to_excel("Course-Schedule.xlsx", index=False)
library(httr)
library(jsonlite)

generate_lecture <- function(seed) {
  body <- list(
    model = "gpt-4o-mini",
    messages = list(
      list(role = "user", content = paste(
        "Generate a lecture title and two-sentence description for:",
        seed
      ))
    )
  )

  res <- POST(
    "https://api.openai.com/v1/chat/completions",
    add_headers(Authorization = paste("Bearer", Sys.getenv("OPENAI_API_KEY"))),
    body = toJSON(body, auto_unbox = TRUE)
  )

  content(res)$choices[[1]]$message$content
}

5.3 Sheet 2 — sections (Delivery Logistics)

5.3.1 Purpose

This sheet captures when and by whom the course is delivered. It is never used to generate .qmd files, but is used for:

  • Syllabus tables
  • Schedule pages
  • Administrative exports

5.3.2 Required Columns

Column Description
Section Section identifier (e.g., A1, B1)
Title Course title
Instructor Instructor name
Day Day of week
Time Start time

5.3.3 Blank Template

Section | Title | Instructor | Day | Time
A1 |  |  |  |

This is intentionally manual-first.

5.4 Sheet 3 — grading (Evaluation Structure)

5.4.1 Purpose

This sheet defines the grading logic independently of content.

It supports:

  • Automatic syllabus grading tables
  • LMS alignment
  • Reweighting without touching content

5.4.2 Required Columns

Column Description
Deliverables Category name
Count_norman Number of items
Total Points Total points allocated

5.4.3 Blank Template

Deliverables | Count_norman | Total Points
Participation |  | 
Weekly Labs |  | 
Individual Assignments |  | 
Group Project Milestones |  |

5.5 Why This Design Works

Principle Outcome
Separation of concerns Content ≠ logistics ≠ grading
AI-optional No forced dependency on OpenAI
Reproducibility Same structure across all courses
Scalability Multiple sections, modalities, instructors
Safety AI never overwrites logistics or grading

5.6 How This Feeds the Pipeline

Sheet Used For
course_details Lecture .qmd, presentations, highlights
sections Schedule tables, syllabus
grading Deliverables page, grading schema

Downstream scripts assume these sheets exist but do not assume they are populated.

5.7 Summary

  • The feeder workbook is the contract
  • AI is assistive, not authoritative
  • Instructors retain full control
  • Automation remains safe and auditable

This structure allows anyone to build a production-grade Quarto course without touching internal code.